Closed exalate-issue-sync[bot] closed 1 year ago
JIRA Issue Migration Info
Jira Issue: PUBDEV-5536 Assignee: New H2O Bugs Reporter: Urs Steinmetz State: Resolved Fix Version: N/A Attachments: Available (Count: 1) Development PRs: N/A
Attachments From Jira
Attachment Name: AE_Grid_Error.R Attached By: Urs Steinmetz File Link:https://h2o-3-jira-github-migration.s3.amazonaws.com/PUBDEV-5536/AE_Grid_Error.R
Since Version 3.18. the h2o.grid search seems to fail for deep learning autoencoder.
Till 3.16. everything is running fine. But in 3.18. the error message "Must specify response, y" occurs. As the Modell is defined to be an autoencoder, this message does not make sense from my point of view. But anyway, if I use almost the same configuration to train a single autoencoder (h2o.deeplearning instead of h2o.grid) erverything runs fine.
The Problem can be traced using the examples attached below.
Example partially taken from: https://github.com/h2oai/h2o-3/blob/master/h2o-docs/src/booklets/v2_2015/source/DeepLearning_Vignette_code_examples/deeplearning_anomaly.R
{code:java} library(h2o) h2o.init(nthreads = -1)
train_ecg <- h2o.importFile( path = "http://h2o-public-test-data.s3.amazonaws.com/smalldata/anomaly/ecg_discord_train.csv", header = FALSE, sep = ",")
test_ecg <- h2o.importFile( path = "http://h2o-public-test-data.s3.amazonaws.com/smalldata/anomaly/ecg_discord_test.csv", header = FALSE, sep = ",")
Layers = list(c(50,15,50), c(50,20,50), c(50,25,50))
anomaly_model <- h2o.deeplearning( x = names(train_ecg), training_frame = train_ecg, validation_frame = test_ecg, activation = "Tanh", autoencoder = TRUE, hidden = c(50,20,50), sparse = TRUE, l1 = 1e-4, epochs = 100)
anomaly_model <- h2o.grid( "deeplearning", hyper_params = Layers,
x = names(train_ecg), training_frame = train_ecg, validation_frame = test_ecg, activation = "Tanh", autoencoder = TRUE, sparse = TRUE, l1 = 1e-4, epochs = 100)
{code}